Daily AI brief

AI news: Astra safeguards, Copilot changes and GLM-5.3-Flash

Four primary-source updates covering a frontier-model safeguard threshold, Copilot model changes and a new Chinese multimodal model.

2026-09-04·Updated 2026-09-04·Primary sources shown for every item

01Policy & safety
OpenAI says Astra reached its critical cybersecurity threshold
OpenAI says Astra is its first model to meet the Critical cybersecurity capability threshold defined in the company’s Preparedness Framework, prompting stronger release safeguards.
Why it mattersCrossing a stated capability threshold makes the safeguards governing model access and release materially important. The evidence reflects OpenAI’s own assessment, not an independent evaluation.
Read analysis
SourcesOpenAI (opens in a new tab)
02Products
GitHub Copilot adds Gemini 3.8 Flash
GitHub has made Google’s Gemini 3.8 Flash available in GitHub Copilot. GitHub says its early testing showed strong results on complex terminal-based coding tasks.
Why it mattersThe addition gives Copilot users another model option for coding workflows, including terminal-oriented tasks, although GitHub’s reported testing has not been independently verified here.
Read analysis
Related directory entriesGemini — Google's assistant, tightly integrated with Search, Gmail and Workspace.GitHub Copilot — The most widely deployed coding assistant, available in most major editors.
SourcesGitHub Changelog (opens in a new tab)
03Products
GitHub schedules Copilot model deprecations for October 2
GitHub says it will deprecate selected models across Copilot Chat, inline edits, ask and agent modes, and code completions on October 2, 2026.
Why it mattersThe change may require users and organizations relying on affected models to update their Copilot configurations or workflows before the deadline. The supplied evidence does not identify the individual models.
Read analysis
Related directory entriesGitHub Copilot — The most widely deployed coding assistant, available in most major editors.
SourcesGitHub Changelog (opens in a new tab)
04Models
Zhipu launches the native multimodal GLM-5.3-Flash model
Zhipu's official BigModel release log lists GLM-5.3-Flash as a newly available native multimodal model, adding another production model option from a Chinese AI provider.
Why it mattersThe release adds to the fast-moving group of Chinese multimodal models available to developers. Its practical position depends on documented capabilities and independent workload testing.
Read analysis
SourcesZhipu BigModel (opens in a new tab)

OpenAI says Astra reached its critical cybersecurity threshold

GitHub Copilot adds Gemini 3.8 Flash

GitHub schedules Copilot model deprecations for October 2

Zhipu launches the native multimodal GLM-5.3-Flash model